Classification of Terahertz Reflection Spectra using Machine Learning Algorithms

A successful implementation of terahertz screening systems requires a development of reliable and efficient identification algorithms. Dimensionality reduction methods are applied to lower the dimensionality of multivariate data while retaining most of the information. Here, we focus on Principal component analysis (PCA) and linear discriminant analysis (LDA) for analysis and classification of terahertz reflection spectra. The complete data set consists of more than 5000 reflection spectra of six active materials. We found that LDA is better for grouping the spectra resulting in highly accurate classification of terahertz spectra. Furthermore, we compare the classification of referenced and non-referenced reflection spectra eligible for real-world applications of terahertz spectroscopy.